WO2022054256A1 - Dispositif de détection d'anomalie - Google Patents

Dispositif de détection d'anomalie Download PDF

Info

Publication number
WO2022054256A1
WO2022054256A1 PCT/JP2020/034581 JP2020034581W WO2022054256A1 WO 2022054256 A1 WO2022054256 A1 WO 2022054256A1 JP 2020034581 W JP2020034581 W JP 2020034581W WO 2022054256 A1 WO2022054256 A1 WO 2022054256A1
Authority
WO
WIPO (PCT)
Prior art keywords
state transition
unit
series data
state
operating state
Prior art date
Application number
PCT/JP2020/034581
Other languages
English (en)
Japanese (ja)
Inventor
敬純 小部
隆彦 増崎
Original Assignee
三菱電機株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 三菱電機株式会社 filed Critical 三菱電機株式会社
Priority to PCT/JP2020/034581 priority Critical patent/WO2022054256A1/fr
Priority to JP2022510985A priority patent/JP7158624B2/ja
Priority to TW109144679A priority patent/TW202210977A/zh
Publication of WO2022054256A1 publication Critical patent/WO2022054256A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring

Definitions

  • This disclosure relates to an abnormality detection device.
  • Patent Document 1 An abnormality is found based on a correlation between physical quantity data acquired from a sensor related to a specific manufacturing machine and operating state data which is information indicating the presence or absence of an abnormality in the operation of the manufacturing machine.
  • An abnormality detector has been proposed that learns a sign and detects a sign of abnormality using this learning result.
  • the estimation result of whether or not the manufacturing machine is operating normally is performed, for example, by outputting the probability that the manufacturing machine is operating normally.
  • the present disclosure has been made to solve such a problem, and an object of the present disclosure is to provide an anomaly detection device capable of performing anomaly detection in consideration of a process flow consisting of a plurality of operating states. do.
  • One aspect of the abnormality detection device is a feature amount extraction unit that receives time-series data of the target device and extracts the feature amount of the time-series data, and an operating state of the target device from the extracted feature amount.
  • the state transition estimation unit that estimates the state transition of the target device from the extracted features by referring to the state transition pattern information that defines the state transition patterns between a plurality of operating states, and the state transition. With reference to the pattern information and the normal range information that defines the normal range for each operating state, the above-mentioned time is determined from the specified operating state, the estimated state transition, and the feature amount of the extracted time-series data. It is provided with an abnormality detection unit for determining whether or not the series data is abnormal.
  • the abnormality determination is made with reference to the state transition pattern information and the normal range information. Therefore, the abnormality determination is performed in two stages of the state transition stage and each operation state stage. be able to. Therefore, it is possible to detect an abnormality in consideration of the flow of a process consisting of a plurality of operating states.
  • FIG. It is a block diagram which shows the structure of the abnormality detection apparatus and the abnormality detection apparatus system by Embodiment 1.
  • FIG. It is a block diagram which shows the detailed structure of the abnormality detection apparatus according to Embodiment 1.
  • FIG. It is a block diagram which shows the hardware configuration example of the abnormality detection apparatus according to Embodiment 1.
  • FIG. It is a block diagram which shows the other hardware configuration example of the abnormality detection apparatus by Embodiment 1.
  • FIG. It is a figure which shows the whole normalization of time series data. It is a figure which shows the example of the division of time series data into each operation state. It is a figure which shows the normalization in each operation state of time series data. It is a figure which shows the example of the calculation of the state transition pattern.
  • FIG. 1 It is a flowchart of the abnormality detection processing of the abnormality detection apparatus in Embodiment 1. It is a block diagram of the learning apparatus for learning the estimation model used by the abnormality detection apparatus of Embodiment 1. FIG. It is a flowchart which shows the learning process of a learning apparatus.
  • FIG. 1 is a block diagram showing a configuration of an abnormality detection device 100 and an abnormality detection system 1000 including the abnormality detection device 100 according to the first embodiment.
  • n target devices n is an integer of 1 or more
  • OD1, OD2, ..., ODn target devices
  • An abnormality detection device 100 that receives time-series data distributed from each of the ODn via the communication network NW, and an external device 200 that accepts various settings by the user or displays the output result of the abnormality detection device 100. It is composed.
  • the target device referred to here may be a single device or a system or device including a plurality of devices in a group.
  • a sensor (not shown) for acquiring data related to the operation of the target device ODn or data related to the operating environment of the target device ODn is arranged in or near the target device ODn.
  • the number of sensors arranged for one target device may be one or may be plural.
  • the plurality of sensors may be of the same type (for example, three accelerometers in the axial direction, the horizontal direction, and the vertical direction), or may be different types (for example, with a voltage sensor). Current sensor).
  • the data to be acquired is not particularly limited as long as it can be acquired as time series data.
  • time-series data examples include current, voltage, power, velocity, acceleration, angular velocity, pressure, magnetic force, torque, temperature, humidity, production, shipments, stock prices, and Internet traffic data.
  • the time-series data may be the detection value of the sensor itself, the statistical value of the detection value (for example, average, maximum, minimum), or the calculated value of the detection value of a plurality of sensors (for example, power).
  • the time-series data acquired from the target device ODn or its vicinity is associated with the identification information of the target device ODn such as the model and the installation location, and is distributed to the abnormality detection device 100 via the wired or wireless communication network NW. ..
  • the abnormality detection device 100 performs state transition estimation and abnormality detection for the target device for which the time series data has been received, based on the time series data received from the target devices OD1, OD2, ..., Or ODn. Further, the abnormality detection device 100 stores the result of the state transition estimation and abnormality detection and the identification information of the target device in the storage unit 4, and when the sensor value has been acquired in response to the input from the external device 200.
  • Series data for example, time-series data of current value, time-series data of Internet traffic data
  • the sensor value at a certain observation time the time-series data of the sensor value over a certain observation period, and the target. Acquire information such as the installation location of the device.
  • the abnormality detection device 100 includes a receiving unit 1, a time-series data analysis unit 2, a data recording control unit 3, a storage unit 4, an interface unit 5, and a time-series data collation unit 6.
  • the receiving unit 1 receives the distribution data D1 distributed from the target devices OD1, OD2, ..., ODn, and from the received distribution data D1, the number of types of sensor values, time-series data of sensor values, sensor items, And the target device information D2 including the usage environment data of the target device is extracted.
  • the receiving unit 1 outputs the extracted target device information D2 to the time series data analysis unit 2 and the data recording control unit 3.
  • the number of types of sensor values means the number of types of sensor data acquired from the sensor mounted on the target device.
  • the type of sensor data is 1.
  • the type of sensor data is two.
  • the number of types shall be counted according to the number of sensors. For example, when data is acquired using three acceleration sensors in the axial direction, the horizontal direction, and the vertical direction, the number of types of sensor data is counted as three.
  • the time-series data of the sensor value means the sensor data in the time-series acquired from the sensor.
  • the sensor item means an item for identifying the sensor mounted on the target device, such as the type of sensor data acquired from the sensor and the installation location of the sensor.
  • the types of sensor data include items related to the operation of the target device such as current, voltage, torque, and temperature, and items related to products manufactured by the target device such as the number of production and the number of shipments.
  • Time series data analysis unit 2 The time-series data analysis unit 2 performs time-series data analysis processing on the target device information D2 received from the reception unit 1.
  • the time-series data analysis unit 2 outputs the analysis result D3 of the target device information D2 to the data recording control unit 3.
  • the analysis result D3 includes the item classification information and the feature amount of the sensor value described later. The details of the time series data analysis unit 2 will be described later.
  • the data recording control unit 3 constructs a database by storing the target device information D2 input from the reception unit 1 and the analysis result D3 input from the time series data analysis unit 2 in the storage unit 4 in association with each other. do.
  • an analysis result D3 (D3-1, D3-2, D3-3, ...) Is generated for each sensor value item.
  • a plurality of analysis results D3 (D3-1, D3-2, D3-3, ...) may be associated with one target device information D2.
  • one target device information D2 includes a voltage value and a current value
  • a voltage value analysis result D3-1 and a current value analysis result D3-2 are generated, and one target device information D2 has a plurality of values.
  • the analysis results D3-1 and D3-2 may be associated with each other.
  • the processing data of a plurality of sensor values may be combined as one analysis result D3, and one analysis result D3 may be associated with one target device information D2.
  • the storage unit 4 stores various data.
  • the storage unit 4 stores the target device information D2 and the analysis result D3 in association with each other. Further, in the first embodiment, the storage unit 4 stores the trained estimation model D4 used by the time-series data collation unit 6 described later.
  • the normal range (normal range information) and the state transition pattern (state transition) of the operating state for each item of a plurality of sensor values associated with the feature amount for each item of a plurality of sensor values are included. Pattern information) is included.
  • the normal range of operating states is defined for each of the plurality of operating states.
  • the normal range of the operating state for example, the range learned by the machine learning method such as regression, Bayesian estimation, confidence interval by statistical method, state space model, 3 ⁇ method, CNN, etc. is used.
  • the user of the abnormality detection device 100 may define a normal range of each operating state.
  • the state transition pattern for example, a range learned by a classification method by machine learning such as causal inference, multivariate analysis, and CNN is used.
  • the state transition pattern of the target device is known in advance from, for example, the specifications of the target device, the user of the abnormality detection device 100 may define the state transition pattern. There may be a plurality of state transition patterns.
  • the storage unit 4 has shown the configuration of storing the target device information D2 and the analysis result D3, but other configurations may be adopted.
  • a single or a plurality of network storage devices (not shown) arranged on the communication network NW store the target device information D2 and the analysis result D3, and the data recording control unit is stored in the network storage device. 3 may be configured to access.
  • the data recording control unit 3 can store the target device information D2 and the analysis result D3 in the external network storage device and build a database outside the abnormality detection device 100.
  • the trained estimation model D4 used by the time-series data collation unit 6 may also be configured to be stored in an external network storage device instead of the storage unit 4.
  • the interface unit 5 connects the external device 200 and each unit of the abnormality detection device 100 to enable communication and various controls.
  • the user of the abnormality detection device 100 can set the monitoring condition D5 in which the monitoring data acquisition unit 61 searches for time-series data by using the external device 200. Further, the user can confirm the collation result D6 of the time series data collation unit 6 by using the external device 200.
  • the time-series data collation unit 6 collates state transition estimation and abnormality detection of time-series data of a plurality of sensor values.
  • the time-series data collation unit 6 acquires the learned estimation model D4 from the storage unit 4, and also obtains the target device information D2 from the storage unit 4 based on the monitoring condition D5 input from the interface unit 5. And the analysis result D3 is acquired, and the sensor value is collated. Further, the time-series data collation unit 6 outputs the sensor value collation result D6 to the interface unit 5, and the collation result D6 is transmitted to the user via the external device 200.
  • FIG. 2 is a configuration diagram showing a detailed configuration of the abnormality detection device 100.
  • the time-series data analysis unit 2 includes a sensor item detection unit 21, a feature amount extraction unit 22, and a sensor item classification unit 23.
  • the sensor item detection unit 21 refers to the data related to the sensor item of the target device information D2 input from the reception unit 1 and the usage environment data of the target device, and the sensor for each usage environment of the target device appearing in the target device information D2. Detect items.
  • the feature amount extraction unit 22 extracts the feature amount of the sensor value for grasping the current operating state and the state transition to the next operating state that can be changed from the current operating state from the time series data of the target device information. do.
  • the feature amount of the sensor value includes the tendency of the waveform of the sensor value in a long time or a short time, the length of the waveform, the frequency, the update frequency of the sensor value, and a plurality of waveforms (time series data). Indicators such as, but not limited to, the similarity between the partial waveforms of.
  • the feature amount of the sensor value may be continuously calculated like a moving average by setting a fixed time width each time the target device information is additionally acquired, or it is calculated by designating an arbitrary time point or interval. You may.
  • the tendency is an index showing the characteristics of the data such as the maximum value or the minimum value, the mean value, the variance or the standard deviation, the correlation, the slope, or the error or the residual.
  • the feature quantity may be extracted from the frequency domain.
  • the sensor item classification unit 23 classifies the feature amount of the extracted sensor value according to the sensor item.
  • the sensor item classification unit 23 is based on the sensor items detected by the sensor item detection unit 21 for each usage environment of the target device and the feature amount of the sensor value extracted by the feature amount extraction unit 22.
  • the feature quantities of the sensor values are classified according to the sensor items.
  • time series data collation unit 6 collates the time-series data of a plurality of sensor values, and determines whether or not the observed value of the time-series data of the sensor values for each operating state is normal.
  • the time-series data collation unit 6 analyzes the abnormality using the learned estimation model D4, and detects the abnormality when the abnormality is detected. The result to that effect is output. Further, when the monitoring condition D5 is set from the external device 200 via the interface unit 5, the analysis result according to the monitoring condition D5 is output.
  • the monitoring condition D5 includes, for example, area information to be monitored, time information to be searched, sensor items to be searched, and feature quantities of sensor values to be searched.
  • the time-series data collation unit 6 includes a monitoring data acquisition unit 61, a processing unit 62, a state transition estimation unit 63, and an abnormality detection unit 64.
  • each component will be described.
  • the monitoring data acquisition unit 61 acquires the target device information D2 of one or a plurality of target devices to be collated, the analysis result D3 associated with the target device information D2, and the learned estimation model D4.
  • the monitoring data acquisition unit 61 stores the sensor items matching the monitoring condition D5 set by the external device 200, the time-series data of the sensor values associated with the sensor items, and the feature amount from the storage unit 4. Search and acquire the corresponding target device information D2 and analysis result D3.
  • the state transition estimation unit 63 associates the target device information D2 and the analysis result D3 acquired by the monitoring data acquisition unit 61 with the learned estimation model D4 acquired from the storage unit 4, and operates the target device during operation. Identify the state and estimate the state transition.
  • the state transition is represented by, for example, the state transition probability, which is the probability of transition from one operating state to another.
  • the state transition estimation unit 63 includes time-series data of sensor values included in the target device information D2, sensor items, usage environment data of the target device, and feature quantities of the sensor values included in the analysis result D3. From, the current operating state (first operating state) is specified.
  • the state transition estimation unit 63 refers to the state transition pattern included in the learned estimation model D4, and uses the feature amount of the sensor value to perform the next operation from the current operation state (first operation state).
  • the state transition probability to the state (second operating state) is calculated.
  • the next operating state (second operating state) that transitions from the current operating state may be one or a plurality.
  • the state transition pattern may be defined by the user of the abnormality detection device 100.
  • the state transition pattern is held as, for example, a state transition table showing a transition relationship between a plurality of operating states and a plurality of operating states.
  • the state transition estimation unit 63 includes the target device information D2 and analysis result D3 acquired from the monitoring data acquisition unit 61, the trained estimation model D4 acquired from the storage unit 4, and the result of processing performed by the state transition estimation unit 63 (the state transition estimation unit 63).
  • the specific result or estimation result) is delivered to the processing unit 62.
  • the target device information D2 and the analysis result D3 may be handed over from the monitoring data acquisition unit 61 to the processing unit 62.
  • the processing unit 62 performs processing performed by the state transition estimation unit 63 such as the input target device information D2, the feature amount of the sensor value included in the analysis result D3 corresponding to the target device information D2, and the probability of the state transition. Get the result and.
  • the processing unit 62 also acquires the trained estimation model D4.
  • the processing unit 62 makes the acquired data or information correspond to the learned estimation model D4, and operates the time-series data of the sensor value of the operating target device appearing in the target device information D2 with a certain probability or more.
  • the time-series data of the state is regarded as each operating state, and normalization, division into operating state units, and calculation of the state transition pattern are performed.
  • the processing unit 62 performs these processes as shown in FIGS. 4A to 4D.
  • the normalization process can be performed on the entire time series data of a series of sensor values as shown in FIG. 4A, or can be performed for each operating state as shown in FIG. 4C. You can also. If there is a difference in the size of the time series data, it may be better to perform normalization, but it is also possible to omit the normalization.
  • Examples of data normalization include, for example, min-max normalization represented by formula (1), z normalization represented by formula (2), and level normalization represented by formula (3). Will be.
  • the min-max normalization converts the range of subsequences from 0 to 1.
  • the z-normalization is a conversion in which the range of the subsequence has an average of 0 and a standard deviation of 1.
  • Level normalization is a transformation in which the mean of subsequences is zero.
  • time-series data as a result of normalizing the time-series data T is expressed as TN .
  • i 1,. .. .. , N
  • the functions min, max, mean, and std are the minimum value, maximum value, average value, and standard deviation of Ti and w , respectively.
  • the time-series data of the sensor values is based on the time-series data of the sensor values. Is divided into each operating state A to D.
  • FIG. 4C when the processing unit 62 is normalized to the time-series data of the sensor values, the processing unit 62 is normalized based on the time-series data of the normalized sensor values.
  • the time-series data of the sensor values are divided into the operating states A to D.
  • the number of divided operating states may be three or less, or five or more.
  • the machining unit 62 calculates the state transition pattern based on the divided operating states A to D.
  • the abnormality detecting unit 64 determines the input target device information D2, the operating state of the sensor value of the operating target device processed by the processing unit 62, and the state transition and the state transition probability estimated by the state transition estimation unit 63. It is used to determine an abnormality in the sensor value of the target device during operation.
  • the abnormality detection unit 64 includes a state transition deviation degree calculation unit 641, an operating state deviation degree calculation unit 642, and a determination unit 643.
  • the state transition deviation degree calculation unit 641 corresponds to the target device information D2 acquired by the monitoring data acquisition unit 61, the state transition and state transition probability estimated by the state transition estimation unit 63, the analysis result D3, and the learned estimation model D4. Then, the degree of deviation from normal in the state transition of the sensor value of the target device in operation is calculated.
  • the state transition deviation degree calculation unit 641 has time-series data of sensor values included in the target device information D2, sensor items, usage environment data of the target device, and a state estimated by the state transition estimation unit 63.
  • the state is used. Calculate the degree of deviation from normal at the transition.
  • the transition probability to the current operating state estimated by the state transition estimation unit 63 (from the operation state immediately before being specified) and the state transition pattern included in the learned estimation model D4 are used. Compare with the transition probability, and use an index that increases as the transition probability of a state transition different from normal increases.
  • the expression "estimated transition probability to the current operating state” includes the meaning that the immediately preceding operating state is determined, but the current operating state is undetermined.
  • the expression "probability of transition to the estimated current operating state” may be expressed as "probability of transition from the specified current operating state to the estimated next operating state".
  • the state transition pattern may be set by the user of the abnormality detection device 100 instead of the learned estimation model D4.
  • the operating state deviation calculation unit 642 includes the target device information D2 acquired by the monitoring data acquisition unit 61, the waveform data and operating state of the sensor value of the operating target device processed by the processing unit 62, and the analysis result D3. By associating the state transition estimated by the state transition estimation unit 63 with the learned estimation model D4, the degree of deviation from the normal operation state of the sensor value of the operating target device is calculated.
  • the operating state deviation calculation unit 642 describes the time series data of the sensor values included in the target device information D2, the sensor items, the usage environment data of the target device, and the sensor values included in the analysis result D3.
  • the degree of deviation from the normal range is compared with the normal range of the waveform included in the trained estimation model D4 corresponding to the operation state processed by the processing unit 62 based on the waveform data sequentially acquired from the detection target. Use an index that increases as the value increases.
  • the normal range of each operating state may be set by the user of the abnormality detection device 100 instead of the learned estimation model D4.
  • the determination unit 643 determines whether or not there is an abnormality in the target device by using the calculation results of the state transition deviation degree calculation unit 641 and the operation state deviation degree calculation unit 642. In the first embodiment, the determination unit 643 determines whether or not there is an abnormality in the target device by using the calculation results of the state transition deviation degree calculation unit 641 and the operating state deviation degree calculation unit 642.
  • the determination method may be calculated as normal or abnormal by performing logical operations such as AND condition and OR condition for each calculation result in the state transition deviation degree calculation unit 641 and the operation state deviation degree calculation unit 642. Alternatively, the degree of abnormality may be calculated numerically by calculating the distance from normal.
  • FIGS. 6A to 6C, and FIGS. 7A to 7C are diagrams showing specific examples of determination by the determination unit 643. It is assumed that the trained estimation model D4 holds the relationship of the state transition diagram of FIG. 5A as a possible state transition for the target device.
  • the state transition diagram of FIG. 5A includes pattern 1 of FIG. 5B and pattern 2 of FIG. 5C as possible state transition patterns for the target device.
  • Pattern 1 is a pattern in which the operating state transitions from the operating state A ⁇ the operating state B ⁇ the operating state C ⁇ the operating state D ⁇ the operating state A.
  • Pattern 2 is a pattern in which the operating state transitions from the operating state A ⁇ the operating state C ⁇ the operating state D ⁇ the operating state A.
  • FIGS. 6A to 6C are diagrams for explaining an abnormality detection example 1 which is an example of an abnormality in a state transition. It is assumed that the time series data represented by the solid line in FIG. 6A is acquired. As a result of the analysis of the time-series data, it was found that there was a state transition from the operating state A to the operating state C, and the time-series data is currently being acquired following the operating state C. In the abnormality detection according to the embodiment of the present disclosure, it is sufficient to know the immediately preceding operating state, so it is sufficient if it is known to be the operating state C, and there is a state transition from the operating state A to the operating state C. It is not necessary to know that it was.
  • the operating state that can be taken after the operating state C is the operating state D.
  • the operating state D is shown by a broken line in FIG. 6A.
  • the time-series data currently being acquired shows a rising edge that follows the waveform of the operating state A. Since the operating state A itself is one of the possible operating states of the target device, the operating state A itself is not abnormal. However, since the operating state A is not an operating state that can be taken after the operating state C in the learned state transition pattern, it can be determined that it is abnormal that the operating state A follows the operating state C. More specifically, it can be determined that the current time-series data following the operating state C is abnormal when a predetermined deviation from the waveform of the operating state D can be confirmed.
  • the state probability of each operating state as shown in FIG. 6B may be used, or the degree of abnormality of the current time series data as shown in FIG. 6C may be used.
  • the probability of a transition that does not correspond to either pattern 1 or pattern 2, which is a pattern of possible state transitions of the target device is the transition to the operating state D. Since it is higher than the probability, it is judged to be abnormal as a transition.
  • FIGS. 7A to 7C are diagrams for explaining an abnormality detection example 2 which is an example of an abnormality in a certain operating state. It is assumed that the time series data represented by the solid line in FIG. 7A is acquired. As a result of the analysis of this time-series data, it was found that the operating state immediately before the current time-series data was the operating state A, and the waveform of the operating state B represented by the broken line and the operating state represented by the broken line are now. The data between the waveforms of C is being acquired. Since the trained estimation model D4 defines a range determined to be normal for each operating state, the time-series data is normal for the range determined to be normal for the waveform of operating state B or the waveform of operating state C.
  • the operating state B or the operating state C is normal. Further, when it is confirmed that the current time-series data in FIG. 7A is not within any of these ranges, it can be determined that the data is abnormal.
  • the state probability of each operating state as shown in FIG. 7B may be used, or the degree of abnormality of the current time series data as shown in FIG. 7C may be used.
  • the state transition is either pattern 1 or pattern 2, which is a pattern of possible state transitions of the target device, but the waveform of the time-series data of the sensor value is pattern 1. And pattern 2 Since it deviates from the normal range of any of the operating states, it is determined that the operating state is abnormal.
  • the abnormality detection unit 64 is specified by the operation state and state transition estimation unit 63 processed by the processing unit 62 with reference to the estimation model D4 that has learned the state transition pattern and the normal range for each operation state. It is determined whether or not the time series data is abnormal from the operation state, the state transition estimated by the state transition estimation unit 63, and the feature amount of the extracted time series data. The state transition probability estimated by the state transition estimation unit 63 may be used. Since the abnormality detection unit 64 makes an abnormality determination by referring to the state transition pattern (state transition pattern information) and the normal range (normal range information) for each operation state, the abnormality determination is performed at the state transition stage and each operation state. It can be done in two stages. Therefore, according to the abnormality detection device 100 according to the first embodiment, it is possible to perform abnormality detection in consideration of the flow of a process composed of a plurality of operating states.
  • the next operating state that transitions from the current operating state is estimated by referring to the learned state transition pattern, and the normal range of the estimated next operating state is estimated by referring to the learned normal range. Is also estimated, so it can be determined that the time series data is abnormal when the time series data deviates from the normal range of the estimated next operating state.
  • the abnormality detection device 100 is realized by, for example, a reception interface 101, an input / output interface 102, and a processing circuit 103.
  • the receiving interface 101 realizes the receiving unit 1
  • the input / output interface 102 realizes the interface unit 5
  • the processing circuit 103 is a time-series data analysis unit 2, a data recording control unit 3, a storage unit 4, and a time-series data collation unit. 6 is realized.
  • the processing circuit 103 executes steps S1 to S9 in the flowchart of FIG.
  • the processing circuit 103 is, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.
  • the functions of the time-series data analysis unit 2, the data recording control unit 3, the storage unit 4, and the time-series data collation unit 6 may be realized by separate processing circuits, and these functions are collectively realized by one processing circuit. You may.
  • the abnormality detection device 100 is realized by, for example, a reception interface 101, an input / output interface 102, a processor 104, and a memory 105.
  • the receiving interface 101 realizes the receiving unit 1
  • the input / output interface 102 realizes the interface unit 5.
  • the program stored in the memory 105 is read out by the processor 104 and executed, the time-series data analysis unit 2, the data recording control unit 3, and the time-series data collation unit 6 are realized.
  • the storage unit 4 is realized by the memory 105.
  • the processor 104 executes steps S1 to S9 in the flowchart of FIG.
  • the program is implemented as software, firmware or a combination of software and firmware.
  • Examples of the memory 105 include, for example, RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically-volatile) or EEPROM (Electrically-EPROM). Includes memory, magnetic disks, flexible disks, optical disks, compact disks, mini disks, and DVDs.
  • time-series data analysis unit 2 data recording control unit 3, storage unit 4, and time-series data collation unit 6 are partially realized by a processing circuit and partly by a processor linked with software or firmware. You may.
  • the data recording control unit 3 is realized by a processing circuit
  • the time-series data analysis unit 2 and the time-series data collation unit 6 are realized by the processor 104 reading and executing the program stored in the memory 105
  • the storage unit. 4 is realized by the memory 105.
  • the functions of the time-series data analysis unit 2, the data recording control unit 3, the storage unit 4, and the time-series data collation unit 6 are realized by a processor that cooperates with a processing circuit, software, or firmware, or a combination thereof. Will be done. It was
  • FIG. 8 is a flowchart showing an abnormality detection process of the abnormality detection device 100 according to the first embodiment.
  • Step S1 the monitoring condition D5 is set by the user of the abnormality detection device 100 via the external device 200.
  • the abnormality detection device 100 acquires the monitoring condition D5 set via the interface unit 5 from the external device 200, and delivers the acquired monitoring condition D5 to the time series data collation unit 6.
  • the monitoring data acquisition unit 61 of the time-series data collation unit 6 determines the target device for abnormality detection based on the received monitoring condition D5. Based on this determination, the abnormality detection device 100 acquires time-series data from the target device ODn that performs abnormality detection, and the feature amount extraction unit 22 extracts and analyzes the feature amount of the sensor value from the time-series data of the target device ODn.
  • the result D3 is output to the recording control unit, and the data recording control unit 3 outputs the analysis result D3 to the storage unit.
  • Step S2 the monitoring data acquisition unit 61 sets the monitoring condition D5 by using the analysis result D3 in the storage unit 4 calculated by the feature amount extraction unit 22 and the learned estimation model D4 in the storage unit 4. Acquire the corresponding feature amount.
  • Step S3 the state transition estimation unit 63 calculates the transition probability for the analysis result D3 based on the monitoring condition D5 based on the feature amount acquired by the monitoring data acquisition unit 61 in step S2.
  • Step S4 the state transition estimation unit 63 calculates the degree of waveform similarity for the analysis result D3 based on the monitoring condition D5 based on the feature amount acquired by the monitoring data acquisition unit 61 in step S2.
  • Step S5 the abnormality detection unit 64 calculates the degree of abnormality of the transition for the analysis result D3 based on the monitoring condition D5 based on the transition probability calculated in step S3.
  • Step S6 the abnormality detection unit 64 calculates the degree of abnormality of the waveform for the analysis result D3 based on the monitoring condition D5 based on the similarity of the waveform calculated in step S4.
  • Step S7 the abnormality detection unit 64 integrates the degree of abnormality of the transition calculated in steps S5 and S6 and the degree of abnormality of the waveform.
  • Step S8 the abnormality detection unit 64 determines whether or not there is an abnormality in the analysis result D3 based on the degree of abnormality integrated in step S7.
  • Step S9 the abnormality detection unit 64 outputs the determination result determined in step S8, and the determination result is displayed on the external device 200 via the interface unit 5.
  • FIG. 6 is a configuration diagram of a learning device 300 for learning the estimation model used by the abnormality detection device 100.
  • the learning device 300 includes a learning data acquisition unit 301 and a model generation unit 302.
  • the learning data acquisition unit 301 acquires the learning data D11 in which the time-series data of the sensor values for each sensor item such as the voltage and current of the target device and the feature amount for each sensor item are associated with each other.
  • the time-series data may be the detection value of the sensor itself, the statistical value of the detection value (for example, average, maximum, minimum), or the calculated value of the detection value of a plurality of sensors (for example, power).
  • Model generation unit 302 Based on the learning data D11 acquired by the learning data acquisition unit 301, the model generation unit 302 has a state transition pattern showing the transition method or order between a plurality of operation states that the target device can take, and each operation. Learn the normal range of time series data in a state.
  • the model generation unit 302 analyzes the time-series data and converts the time-series data (hereinafter, may be referred to as “waveform”) into a plurality of waveforms (hereinafter, referred to as “waveforms”) corresponding to a plurality of operating states. It is sometimes called "partial waveform").
  • the operating state means a state of operation that the target device can take when the operation of the target device is broadly classified.
  • the operating state of the actuator is, for example, a state in which the current driving the actuator rises, continues to peak, and falls.
  • the analysis of the time series data is performed, for example, by acquiring event data indicating the mutation point of the waveform of the data.
  • the Ramer-Douglas-Pucker (RDP) algorithm can be used to detect the mutation point.
  • a method such as autocorrelation or dynamic time warming (DTW), multivariate analysis such as principal component analysis or discriminant analysis, or an estimation method such as a support vector machine (SVM) may be used. ..
  • DTW dynamic time warming
  • SVM support vector machine
  • event data indicating the variation point of the operating state may be acquired, or causal inference, multivariate analysis, classification method by machine learning such as CNN, etc. may be used. Further, the appearance order of the operating states may be visualized and presented to the user of the abnormality detection device 100, and the user may define the state transition pattern.
  • the model generation unit 302 first identifies the operating state of the divided partial waveform (for example, A, B, C, D, etc.). ) Is assigned.
  • the operating state ID can be regarded as a clustering problem. Since it is conceivable that the operation order of the manufacturing machine is not fixed and the operation follows a plurality of state transition patterns, clustering may be performed for each operation state, that is, for each partial waveform shape.
  • the k-means method which is a method of unsupervised learning, can be used. By using this method, partial waveforms can be clustered for each similar shape, and an operation state ID can be assigned for each similar shape.
  • the normal range for each operating state regression, confidence intervals by Bayesian inference or statistical methods, state-space models, and machine learning methods such as the 3 ⁇ method and CNN may be used.
  • the normal range is R1 for one operating state 1, the normal range is R2 for another operating state 2, and R1 is narrower than R2.
  • the abnormalities can be appropriately determined for the operating state 1, but the abnormalities cannot be appropriately determined for the operating state 2. This is because the range determined to be normal is set narrower than the original R2 for the operating state 2 (the normal range is too small for the operating state 2). Therefore, the operation state 2 may be determined to be abnormal even though the operation is normal.
  • the abnormality can be appropriately determined for the operating state 2, but the abnormality cannot be appropriately determined for the operating state 1.
  • the range determined to be normal for the operating state 1 is set wider than that for the original R1 (the normal range is excessive for the operating state 1). Therefore, the operation state 1 may be determined to be normal even though the operation is abnormal.
  • the learning device 300 is used to learn the learning model used by the abnormality detection device 100.
  • the learning device 300 is connected to the abnormality detection device 100 via a network and is a device separate from the abnormality detection device 100. There may be. Alternatively, the learning device 300 may be built in the abnormality detection device 100.
  • the model generation unit 302 obtains the trained estimation model D12 which is the learning result of the state transition pattern (state transition pattern information) and the normal range (normal range information) for each operating state. It is generated and output to the storage unit 4 of the abnormality detection device 100.
  • the storage unit 4 stores the trained estimation model D12 output from the model generation unit 302.
  • FIG. 10 is a flowchart showing the learning process of the learning device 300.
  • step S21 the learning data acquisition unit 301 acquires the learning data D11 in which the time-series data of the sensor value for each sensor item of the target device and the feature amount for each sensor item are associated with each other.
  • step S22 the model generation unit 302 analyzes the time-series data of the sensor values for each sensor item of the target device based on the learning data D11 acquired by the learning data acquisition unit 301, and a plurality of time-series data. It is divided into the operating states of, and the state transition pattern between the divided operating states and the normal range of the time series data in each operating state are learned to generate an estimation model.
  • step S23 the storage unit 4 stores the trained estimation model D12 generated by the model generation unit 302.
  • the time-series data collated by the time-series data collation unit 6 may be data acquired in real time or may be data acquired in the past. Further, abnormality detection may be performed on all the acquired time-series data of the sensor values.
  • the abnormality detection device (100) receives the time-series data of the target device and extracts the feature amount of the time-series data, and the feature amount extraction unit (22) and the operating state of the target device from the extracted feature amount.
  • the state transition estimation unit (63) that estimates the state transition of the target device from the extracted features by referring to the state transition pattern information that defines the state transition patterns between the plurality of operating states. With reference to the state transition pattern information and the normal range information defining the normal range for each operating state, from the specified operating state, the estimated state transition, and the feature quantity of the extracted time series data.
  • an abnormality detecting unit (64) for determining whether or not the time series data is abnormal.
  • the abnormality detection device of Appendix 2 is the abnormality detection device described in Appendix 1, and the state transition estimation unit estimates the state transition by calculating the state transition probability.
  • the abnormality detection device of the appendix 3 is the abnormality detection device described in the appendix 2, and the abnormality detection unit acquires an operation state transitioning from the specified operation state with reference to the state transition pattern information. Then, when the state transition probability of the estimated state transition becomes higher than the probability of transitioning to the acquired operating state, it is determined that the estimated state transition is abnormal.
  • the abnormality detection device of the appendix 4 is the abnormality detection device described in any one of the appendices 1 to 3, and the abnormality detection unit refers to the normal range information and is in the specified operating state. When the normal range is acquired and the waveform corresponding to the specified operating state of the time series data deviates from the normal range, it is determined that the time series data is abnormal as waveform data.
  • the abnormality detection device of the appendix 5 is the abnormality detection device described in any one of the appendices 1 to 4, and the abnormality detection unit deviates from the normal state transition of the estimated state transition.
  • the abnormality detection device of the appendix 6 is the abnormality detection device described in any one of the appendices 1 to 5, and the state transition pattern information and the normal range information are obtained as a trained estimation model by machine learning. Information.
  • the abnormality detection device of the appendix 7 is the abnormality detection device described in the appendix 6, further including a learning device (300) for learning the estimation model, and the learning device is a state from the feature amount of the time series data.
  • a model generation unit (302) for calculating a transition pattern and a normal range for each operating state is provided.
  • the abnormality detection device of the present disclosure can detect an abnormality in time-series data, for example, a device for preventive maintenance of a production system in a factory or a plant, a system used in social infrastructure such as a railroad or a station. It can be used as a device for preventive maintenance of devices and devices, and as a monitoring device for economic indicators such as stock prices.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

Le dispositif de détection d'anomalie d'après la présente invention comprend : une unité d'extraction de quantité caractéristique (22) qui reçoit des données de série chronologique d'un dispositif cible et extrait une quantité caractéristique des données de série chronologique ; une unité d'estimation de transition d'état (63) qui identifie l'état de fonctionnement du dispositif cible à partir de la quantité caractéristique extraite, se réfère à des informations de modèles de transition d'état spécifiant des modèles de transition d'état entre une pluralité d'états de fonctionnement et estime une transition d'état du dispositif cible à partir de la quantité caractéristique extraite ; et une unité de détection d'anomalie (64) qui se réfère aux informations de modèles de transition d'état et à des informations sur la plage normale spécifiant la plage normale pour chaque état de fonctionnement et détermine si les données de série chronologique sont anormales à partir de l'état de fonctionnement identifié, de la transition d'état estimée et de la quantité caractéristique extraite des données de série chronologique.
PCT/JP2020/034581 2020-09-11 2020-09-11 Dispositif de détection d'anomalie WO2022054256A1 (fr)

Priority Applications (3)

Application Number Priority Date Filing Date Title
PCT/JP2020/034581 WO2022054256A1 (fr) 2020-09-11 2020-09-11 Dispositif de détection d'anomalie
JP2022510985A JP7158624B2 (ja) 2020-09-11 2020-09-11 異常検知装置
TW109144679A TW202210977A (zh) 2020-09-11 2020-12-17 異常檢知裝置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2020/034581 WO2022054256A1 (fr) 2020-09-11 2020-09-11 Dispositif de détection d'anomalie

Publications (1)

Publication Number Publication Date
WO2022054256A1 true WO2022054256A1 (fr) 2022-03-17

Family

ID=80631449

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2020/034581 WO2022054256A1 (fr) 2020-09-11 2020-09-11 Dispositif de détection d'anomalie

Country Status (3)

Country Link
JP (1) JP7158624B2 (fr)
TW (1) TW202210977A (fr)
WO (1) WO2022054256A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023218550A1 (fr) * 2022-05-11 2023-11-16 三菱電機株式会社 Équipement de correction, procédé de traitement, et programme de traitement
WO2024047694A1 (fr) * 2022-08-29 2024-03-07 三菱電機株式会社 Dispositif d'aide au fonctionnement, système d'aide au fonctionnement et procédé d'aide au fonctionnement

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015145865A1 (fr) * 2014-03-24 2015-10-01 日本電気株式会社 Dispositif de surveillance, système de surveillance, procédé de surveillance et programme
JP2019204155A (ja) * 2018-05-21 2019-11-28 ファナック株式会社 異常検出器
WO2019244203A1 (fr) * 2018-06-18 2019-12-26 三菱電機株式会社 Dispositif et procédé de diagnostic, et programme

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2752722B1 (fr) 2011-08-31 2019-11-06 Hitachi Power Solutions Co., Ltd. Procédé de surveillance d'état d'installation et dispositif pour celui-ci
US10712733B2 (en) 2016-12-12 2020-07-14 Mitsubishi Electric Research Laboratories, Inc. Methods and systems for discovery of prognostic subsequences in time series

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015145865A1 (fr) * 2014-03-24 2015-10-01 日本電気株式会社 Dispositif de surveillance, système de surveillance, procédé de surveillance et programme
JP2019204155A (ja) * 2018-05-21 2019-11-28 ファナック株式会社 異常検出器
WO2019244203A1 (fr) * 2018-06-18 2019-12-26 三菱電機株式会社 Dispositif et procédé de diagnostic, et programme

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023218550A1 (fr) * 2022-05-11 2023-11-16 三菱電機株式会社 Équipement de correction, procédé de traitement, et programme de traitement
WO2024047694A1 (fr) * 2022-08-29 2024-03-07 三菱電機株式会社 Dispositif d'aide au fonctionnement, système d'aide au fonctionnement et procédé d'aide au fonctionnement

Also Published As

Publication number Publication date
TW202210977A (zh) 2022-03-16
JPWO2022054256A1 (fr) 2022-03-17
JP7158624B2 (ja) 2022-10-21

Similar Documents

Publication Publication Date Title
EP2905665B1 (fr) Appareil de traitement d'informations, procédé de diagnostic et programme
US20180239345A1 (en) Abnormality predictor diagnosis system and abnormality predictor diagnosis method
EP3186514B1 (fr) Surveillance de condition de pompe
WO2012032812A1 (fr) Procédé de détection de défauts de fonctionnement et système correspondant
EP3552067B1 (fr) Procédés et systèmes de découverte de sous-séquences de pronostic dans une série chronologique
CN110515351A (zh) 异常检测器
WO2011086805A1 (fr) Procédé de détection d'anomalie et système de détection d'anomalie
US20070239629A1 (en) Cluster Trending Method for Abnormal Events Detection
WO2022054256A1 (fr) Dispositif de détection d'anomalie
JP2004531815A (ja) 予測的状態監視のための診断システムおよび方法
US20210125323A1 (en) System and method for determining situation of facility by imaging sensing data of facility
CN107729985B (zh) 识别技术设施的过程异常的方法和相应的诊断***
JP2014149840A (ja) 異常検知方法及びそのシステム
CN105675038A (zh) 一种仪表的故障预测装置
US20210310679A1 (en) Abnormality cause estimation device, abnormality cause estimation method, and medium
KR20190082715A (ko) 상관도를 고려한 데이터 분류 방법 및 이 방법을 수행하기 위한 프로그램이 저장된 컴퓨터 판독가능한 저장매체
CN117076869B (zh) 旋转机械时频域融合故障诊断方法及***
KR101997580B1 (ko) 상관도를 고려한 데이터 분류 방법 및 이 방법을 수행하기 위한 프로그램이 저장된 컴퓨터 판독가능한 저장매체
WO2020230422A1 (fr) Dispositif de diagnostic d'anomalie et procédé
US20230251167A1 (en) Learning device, defect detection device, and defect detection method
US20220405161A1 (en) Data selection assist device and data selection assist method
EP4033219B1 (fr) Dispositif de détermination d'anomalie et procédé de détermination d'anomalie
US20230102000A1 (en) Time-series pattern explanatory information generating apparatus
Yu et al. Novelty class detection in machine learning-based condition diagnosis
US11954131B2 (en) Time-series data processing method

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2022510985

Country of ref document: JP

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20953321

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20953321

Country of ref document: EP

Kind code of ref document: A1